In the context of digital twins, surrogate model reconstruction is essential for real-time control, fault diagnosis, and performance prediction. While data-driven techniques are widely used, their accuracy heavily depends on the availability of large, high-quality datasets. In contrast, projection-based model order reduction (MOR) techniques, such as Proper Orthogonal Decomposition (POD), preserve the structure of the physical model, ensuring greater interpretability. However, these methods often lack robustness when handling multiple varying parameters, which can compromise their accuracy and reliability in multi-parameter electromagnetic problems. To address these challenges, a hybrid approach leveraging tensor decomposition has recently been proposed. This method enables the identification of an optimal low-dimensional representation of the parametric problem, thereby enhancing the robustness of surrogate models in both data-driven and model-driven techniques. Applied to multi-parameter electromagnetic problems across various scenarios, the approach has demonstrated superior performance compared to classical methods. These results highlight its potential for advancing digital twin applications in electrical engineering.